rwalk mcmc hanging?
During the 150914 pbilby tests, I've noticed that when running on 128 and 256 cores with 1000 walks, the sampler seems to hang when the mcmc kicks in. Roughly, 20-something minutes are spent going from the first to the second rwalk call. After this period of hanging, the run seems to proceed smoothly again. Why does the sampler hang? Maybe communication blocking? Or perhaps an issue with the cluster?